SIRE-BY-HERD INTERACTION EFFECT WHEN VARIANCES ACROSS HERDS ARE HETEROGENEOUS .1. EXPECTED GENETIC PROGRESS

Six different variance structures were simulated across herds to fit a real data structure with 197 335 lactations in 924 herds. The homogen eity case, as well as different combinations of heterogeneous variance s across herd-year-seasons (HYSs) were considered. A positive correlat ion of 0.4 was assumed between average production and phenotypic stand ard deviation within HYS. Two animal models, with and without sire-by- herd (S x H) interaction, were compared in their ability to rank the b est animals. The interaction term was used as a correction factor for heterogeneity. In terms of accuracy of genetic proofs and estimated ge netic progress, the model including the S x H interaction effect resul ted in less accurate genetic proofs compared to a model without S x H interaction. Decrease in accuracy was greater for sires (-0.03) than f or cows (-0.01). The estimated genetic progress was 2-3 % lower when f itting the interaction into the model than when not considering S x H effects.